Whisper with Timestamps

Generate audio timestamps for your readalong project. This walks through installing OpenAI's Whisper in an isolated environment on Ubuntu, running it against an MP3, and pulling out word-level timestamps as JSON.

#01Install dependencies

First, we need to install Python virtual environment support and ffmpeg for audio processing.

sudo apt install python3-venv python3-full ffmpeg

#02Create virtual environment

Create an isolated Python environment to keep Whisper and its dependencies separate from your system Python.

python3 -m venv ~/whisper-env

#03Activate virtual environment

Activate the environment to start using it. You'll see (whisper-env) in your terminal prompt.

source ~/whisper-env/bin/activate

#04Install Whisper

Now install OpenAI's Whisper package inside your virtual environment.

pip install -U openai-whisper

#05Run Whisper on your audio

Execute Whisper to generate timestamps from your MP3 file.

whisper /home/youruser/7R3M0R.mp3 --model base --output_format json --word_timestamps True --output_dir /home/youruser/whisper_output

#Available Models

#06Understanding Whisper output

During processing, Whisper will detect the language and show progress with timestamps. Here's what you'll see:

Detecting language using up to the first 30 seconds. Use `--language` to specify the language
Detected language: English
[00:00.000 --> 00:00.600]  Tremor.
[00:01.720 --> 00:08.660]  Above the smoldering ruins of Earth, hidden from the Sun's wrath, the USS drifted silently.
[00:09.520 --> 00:16.180]  It wasn't the old station. This was its successor, the United Space Station, commissioned
[00:16.180 --> 00:20.780]  by the US government and built by the one mind they feared yet still couldn't live without.
[00:21.660 --> 00:28.360]  Tremor. His triumph with truth.ai had sealed his legend, not just as a visionary,
[00:28.920 --> 00:33.440]  but as a ruthless, brilliant tech magnate. And he delivered.

#07Retrieve your files

Once processing is complete, find your output files in the directory.

ls /home/youruser/whisper_output/

Whisper generates multiple output formats:

• 7R3M0R.json — JSON with detailed timestamp data
• 7R3M0R.txt — Plain text transcription
• 7R3M0R.srt — Subtitle format (optional)
• 7R3M0R.vtt — WebVTT format (optional)

cat /home/youruser/whisper_output/7R3M0R.json

The JSON file contains word-level timestamps perfect for creating your readalong!

#Next Steps

Now that you have your timestamp file, you can:

• Upload the JSON file along with your novel chapter text
• Create an interactive readalong with synchronized highlighting
• Build a custom web interface for your audiobook

Ready to build your readalong? Share your JSON file and chapter text!

#Future Use

When you're done working with Whisper, deactivate the virtual environment:

deactivate

To use Whisper again in the future, just activate the environment:

source ~/whisper-env/bin/activate